TWI627536B - System and method for a shared cache with adaptive partitioning - Google Patents

Abstract

The present invention provides a cache controller that adaptively divides a shared cache. The adaptive split cache controller includes flag comparison and obsolete logic and selection logic in response to client access requests and various parameters. A component cache is assigned a target occupancy count that is compared to a current occupancy count. Conditional identification of one of the obsolete cache lines to manage the data stored in the shared cache. When a conflict or cache miss is identified, the selection logic identifies candidates that are preferably replaced between cache lines that are identified as obsolete. For a fixed number of banks per component cache, each cache line is assigned to a bank. The assigned cache line is assigned to a bank based on the target occupancy. Subsequent allocations make the oldest cache line obsolete after the selected number of banks have been populated. The cache lines are considered obsolete when the individual component cached indicator of the cache line is revoked.

Description

System and method for sharing cache with one of adaptive segmentation

Cache segmentation to serve multiple client programs or client applications ("user terminals") in a computing system has several benefits. The partition size can be varied to match the performance requirements of the client. The client experiences robust performance because the data in its private cache is not subject to retraction due to access initiated by other clients. However, such benefits are not as large as they might be due to the limitations imposed by the way the current cache is split. Some conventional designs of computing devices include multiple processors of different types (such as central processing units, graphics processing units, display controllers, hardware accelerators, etc.) and/or having multiple cores to support the various types of computing devices required for a particular computing device. Main and peripheral functions of the processor. Such designs often further integrate analog, digital, and RF circuits or other specialized circuits to achieve a particular function on a single germanium substrate, and are commonly referred to as system single-chip (SoC). Currently, the cache can be split across only the cache path or a combination of both path split and set split, all of which are limited. When only path segmentation is implemented, the number of cache partitions is limited by the number of cache paths, and adding associations to the cache partitions to reduce cache collisions reduces the maximum number of possible cache partitions. The minimum path partition size is determined by the association. For example, the smallest partition of an 8 MB cache with 8 lanes (ie, 8 lane associations) is 1 MB. Also, enabling the use of smaller cache partitions may cause loss of cache space. Using the same example, 512 KB of data stored in a minimum 1 MB cache partition will produce 512 KB of unused space in the cache partition. Set partitioning maintains high correlation and provides greater flexibility, but creates data consistency management issues that may require complex solutions for successful management because it is not a trivial task to resize the cache in the collection size. The collection split works in conjunction with the baseline collection association cache. However, set splitting does not work with skewed associative cache structures. Path segmentation is illustrated in FIG. The cache storage area 1 includes an array of addressable memory segments that can be individually addressed by N lanes and M sets, where N and M are positive integers. As further indicated in FIG. 1, the cache storage area 1 includes 8 lanes and 8 sets, and is segmented into cache partitions 2, 3, and 4, wherein the cache partition 2 includes 4×8 addressable sections. The capacity of the cache partition 3 includes 2 x 8 addressable segments, and the cache partition 4 includes 2 x 8 addressable segments. The set and path segmentation are illustrated in FIG. The cache storage area 5 includes an array of addressable memory segments that can be individually addressed by N lanes and M sets, where N and M are positive integers. As further indicated in FIG. 2, the cache storage area 5 includes 8 lanes and 8 sets, and is segmented into cache partitions 6, 7, and 8, wherein the cache partition 6 includes 4×4 addressable sections. The capacity of the cache partition 7 includes 2 x 1 addressable sectors, and the cache partition 8 includes 1 x 8 addressable sectors. The effectiveness of the combined set/path segmentation scheme is questionable, which exceeds a relatively small number of coarse-grained and highly customized solutions. When generating and deleting component caches and resizing them, it is necessary to manipulate and manage the complex geometric representation of the aggregate/path segmentation cache over time. Therefore, the set/path segmentation method is not scalable. Therefore, there is a need for an improved mechanism for dynamically managing cache partitions.

An alternative embodiment for a computing system and method for adaptively segmenting a shared cache is disclosed. Example embodiments illustrate hardware components, including storage registers, interconnects, and logic circuitry that can be deployed in an improved cache controller in a computing device.

Another example embodiment includes a computing device. The computing device includes a processor, a shared cache, and a shared or adaptive split cache controller. The shared cache provides a data storage resource for the processor. The adaptive split cache controller is communicatively coupled to the processor and the shared cache. The adaptive segmentation cache controller includes an input port, a register, a tag comparison and outdated logic module per channel, and a selection logic module. The input terminal receives an access request from a client executing on or communicating with the computer device. The registers provide data storage for operational parameters for a desired number of component caches or segments of the shared cache. The tag comparison and obsolescence logic is coupled to the registers and configured to receive one of the elapsed times for which information has been stored in a particular cache line of the shared cache, to record an obsolete condition and conditional Update a storage indicator. The selection logic is coupled to the tag comparison and obsolete logic and the registers. The selection logic is configured to conditionally identify one of the replacement candidates to be replaced in the shared cache.

Another example embodiment is a method for adaptively splitting a shared cache. The method includes the steps of: storing parameters defining an integer number of component caches in the shared cache, wherein at least one of the integer number of component caches includes a target occupancy number and an occupancy count One set of parameters is associated; a unique identifier corresponding to one component cache of the integer number of component caches is applied to the client request; in response to the client storing the client data in the shared cache Requesting, using information stored in the set of parameters for cache line allocation, wherein when a cache line is allocated, incrementing an associated occupancy count; and using the information stored in the set of parameters to identify Retrieved candidate cache line, where when reclaiming a cache line When the associated occupancy count is decremented.

In another embodiment, an adaptive segmentation controller includes a scratchpad, tag comparison and obsolescence logic, and selection logic. The registers store operational parameters of a desired number of component caches within a shared cache. The tag comparison and obsolescence logic is coupled to the registers, and receives a cache address and a unique component cache identifier identifying one of the cache lines of the shared cache. In response to an operation mode input, the tag comparison and obsolete logic generates a measure of the elapsed time that the information has been stored in a particular cache line of the shared cache, conditionally records an obsolete condition, a valid condition, and Update a storage indicator. The selection logic is coupled to the tag comparison and obsolete logic and the registers. The selection logic is configured to receive the measure of elapsed time for the information stored in the particular cache line in the shared cache, the unique component cache identifier, and a priority indicator. The selection logic identifies one of the candidate cache lines to be replaced in the shared cache based on the obsolete condition, a valid condition, and the component cache identifier.

In an alternative embodiment, portions of the novel cache controller and method may be implemented by a non-transitory processor readable medium having stored thereon a processor that, when executed, directs a processor And a processor-executable instruction that adaptively splits the cache controller to perform operations. The operations include: defining configuration parameters, occupancy parameters, and time-related parameters, including a unique identifier, a target occupancy number, an occupancy count, and an integer number of storage indicators; identifying a share from a client application Cache access request; generate one of the elapsed times for which information has been stored in a particular cache line in a system cache; record an obsolete condition; conditionally update a storage index; and conditionally identify One of the candidate candidates to be replaced in the system cache.

The word "exemplary" is used herein to mean "serving as an example, instance, or illustration." Any aspect described herein as "exemplary" is not necessarily to be construed as preferred or advantageous. In this description, the term "application" may also include files having executable content such as object code, instruction code, byte code, markup language file, and patch. In addition, the "applications" mentioned herein may also include files that are not executable in nature, such as documents that may need to be opened or other data files that need to be accessed. Various aspects will be described in detail with reference to the accompanying drawings. Wherever possible, the same reference numerals reference The specific examples and implementations are for illustrative purposes and are not intended to limit the scope of the systems and methods of the invention or the scope of the claims. The terms "computing device" and "mobile device" are used interchangeably herein to refer to any or all of the following: a cellular telephone, a smart telephone, a personal or mobile multimedia player, a personal data assistant (PDA) ), laptops, tablets, smart laptops, ultrabooks, palmtops, wireless email receivers, cellular phones with multimedia Internet capabilities, wireless game controllers and similar individuals including memory Electronic devices, and programmable processors. While the various aspects are particularly applicable to mobile computing devices, such as smart phones, which have limited energy resources, the various aspects are generally applicable to any electronic device that implements a plurality of memory devices and in which energy efficiency is a concern. Thus, the adaptive segmentation cache controller and method can be applied to computing devices configured in a desktop and rack-mounted form factor, and the computing device typically only uses the battery when other power sources are not available. The term "system-on-a-chip" (SoC) is used herein to refer to an electronic circuit that typically (but not exclusively) includes a group of hardware cores, memory, and communication interfaces. The hardware core may include a plurality of different types of processors, such as general purpose processors, central processing units (CPUs), digital signal processors (DSPs), graphics processing units (GPUs), accelerated processing units (APUs), auxiliary processors, Single core processor and multi-core processor. The hardware core can further embody other hardware and hardware combinations, such as field programmable gate array (FPGA), special application integrated circuits (ASIC), other programmable logic devices, discrete gate logic, transistor logic, performance Monitor hardware, watchdog hardware, and time reference. The integrated circuit can be configured such that components of the integrated circuit reside on a single piece of semiconductor material, such as germanium. The term "content" is used herein to describe a file having executable content, such as: object code, instruction code, byte code, markup language file, and patch. In addition, the "content" mentioned in this document may also include files that are not executable in nature, such as data files or data values that need to be accessed. As used in this specification, the terms "component," "module," "system," and the like are intended to refer to a computer-related entity that is a combination of hardware, firmware, hardware, and software, software, or implementation. Any of the software. For example, a component can be, but is not limited to being, a processor, a processor, an object, an executable, a thread, a program, and/or a computer executing on a processor. By way of illustration, both an application executing on a computing device and the computing device can be a component. One or more components can reside within a process and/or thread, and a component can be located on a computer and/or distributed between two or more than two computers. In addition, such components can be executed from a variety of processor readable media having various data structures stored thereon. The component may, for example, be based on a signal having one or more data packets (eg, from the signal with the other system in the local system, the distributed system, and/or across the network (such as the Internet) The data of one component of other system interactions) communicates by means of the local and/or remote processing program. As used in this specification, the term "adaptation" describes and includes that it can grow beyond the target occupancy when there is no parallel between multiple clients, but stay close to the target if there is parallel use between multiple clients. The cache of the number of occupations. The adaptive split cache controller will guarantee the minimum occupancy of each component cache, regardless of relative parallelism. In addition, component caches can take up additional storage space opportunistically when parallelism is low, resulting in increased performance. Thus, in some embodiments, the adaptive segmentation cache controller can achieve better performance by dynamically adjusting the software overhead and intervention of the absence of component cache parameters. An example embodiment includes an adaptive split cache controller. The adaptive split cache controller includes a scratchpad, a tag comparison and an outdated logic module, and a victim selection logic module. There is one tag comparison and outdated logic per channel. The registers are configured to store a set of operational parameters of a desired number of component caches of the shared cache. Each state logic module is communicatively coupled to the scratchpad and configured to receive a cache address associated with the transaction presented to the cache, a unique component cache identifier, and a particular attribute selected as appropriate. Each of the tag comparison and obsolete logic modules generates, in response to the current bank indicator, the overflow bit, the active bit, the priority level, and the replacement strategy mode, the information is stored in a particular cache line in the shared cache. The time indicator is measured, the obsolete condition is recorded, and the storage indicator is updated according to the replacement strategy mode. Under certain conditions, the identified component cache is allowed to exceed the number of target assignments assigned by each. The victim selection logic is communicatively coupled to the tag comparison and obsolete logic module and the scratchpad. The victim selection logic is configured to receive, in parallel from each of the tag comparison and obsolete logic modules, a measure of elapsed time, component cache identifier, priority level, obsolete, for information stored in a particular cache line in the shared cache. And valid conditions, and the candidate cache lines to be replaced in the shared cache are identified according to the victim selection algorithm. The selection algorithm will preferentially select non-valid candidates, otherwise select outdated candidates, otherwise select lower priority candidates, otherwise select the oldest candidate with the same component cache identifier, otherwise the victim will not be selected.

3 illustrates a system that includes computing device 10 in communication with a remote computer 50 suitable for use with a variety of aspects. The computing device 10 includes a SoC 12 having a processor 14, a memory 16, a communication interface 18, and a storage interface 20. The computing device 10 further includes a communication component 22, such as a wired or wireless data modem, a storage component 24, an antenna 26 for establishing a wireless connection 32 to the wireless network 30, and/or a wired connection for connecting to the Internet 40. 44 network interface 28. Processor 14 can include any of a variety of hardware cores as well as several processor cores. Computing device 10 may also include one or more processors 14 and two or more SoCs 12, thereby increasing the number of processors 14 and processor cores. Computing device 10 may include a separate processor core 14 that is not associated with SoC 12. The individual processor 14 can be a multi-core processor as described below with reference to FIG. Processors 14 may each be configured to be the same or different than other processors 14 of computing device 10 for a particular purpose. One or more of processor 14 and processor cores having the same or different configurations may be grouped together as part of one or more subsystems of computing device 10 as described below with reference to FIG.

The memory 16 of the SoC 12 can be a volatile or non-volatile memory configured to store data and processor executable code for access by the processor 14. In one aspect, the memory 16 can be configured to at least temporarily store a data structure, such as a table of component caches for managing cached memory that is adaptively segmented as described below. As discussed in further detail below, each of the processor cores can access a plurality of component caches of the cache memory.

Computing device 10 and/or SoC 12 may include one or more memories 16 configured for various purposes. In one aspect, one or more of the memory 16 can be configured to be dedicated to store data for storage by enabling the data structure information to be accessed by the processor 14 for managing component cache access requests. Take the information structure of the information. When the memory 16 storing the data structure is non-volatile, the memory 16 can maintain information on the data structure even after the power of the computing device 10 has been turned off. Information about the data structure stored in non-volatile memory 16 may be available to computing device 10 when power is restored and computing device 10 is restarted. Communication interface 18, communication component 22, antenna 26, and/or network interface 28 may work in conjunction to enable computing device 10 to connect via 44 and/or wired network connection 44 and the Internet including remote computing device 50 The nodes on 40 communicate via wireless network 30. The wireless network 30 can be implemented using a variety of wireless communication technologies including, for example, a radio frequency spectrum for wireless communication to provide a computing device 10 coupled to the Internet 40, which can be computed over the Internet and remotely Device 50 exchanges data. The storage interface 20 and the storage component 24 can cooperate to allow the computing device 10 to store data on a non-volatile storage medium. The storage component 24 can be configured to closely resemble one aspect of the memory 16, wherein the storage component 24 can store the data structure such that information stored in the data structure can be accessed by the one or more processors 14. The stored information is maintained for the non-volatile storage component 24 even after the power to the computing device 10 has been turned off. The information stored in storage component 24 can be made available to computing device 10 when power is restored and computing device 10 is restarted. The storage interface 20 controls access to the storage component 24 and allows the processor 14 to read data from the storage component 24 and write data to the storage component 24. Some or all of the components of computing device 10 may be configured and/or combined in different ways while still providing the necessary functionality. Moreover, computing device 10 may not be limited to one of each of the components, and multiple instances of each component may be included in various configurations of computing device 10. FIG. 4 illustrates a multi-core processor 14 suitable for implementing an aspect. Multi-core processor 14 may have a plurality of equivalent processor cores 200, 201, 202, 203. Processor cores 200, 201, 202, 203 may be equivalent, as processor cores 200, 201, 202, 203 of a single processor 14 may be configured for the same purpose and have the same or similar performance characteristics. For example, processor 14 can be a general purpose processor, and processor cores 200, 201, 202, 203 can be equivalent general purpose processor cores. Alternatively, processor 14 can be a GPU or a DSP, and processor cores 200, 201, 202, 203 can each be an equivalent GPU core or a DSP core. The performance characteristics of the processor cores 200, 201, 202, 203 may be within the same multi-core processor 14 or within another multi-core processor 14 using a processor core of the same design, via variations in manufacturing processes and materials. The core is different. In the example illustrated in FIG. 4, multi-core processor 14 includes four processor cores 200, 201, 202, 203 (ie, processor core 0, processor core 1, processor core 2, and processor core). 3). For ease of explanation, the examples herein may refer to the four processor cores 200, 201, 202, 203 illustrated in FIG. It should be noted, however, that the four processor cores 200, 201, 202, 203 illustrated in FIG. 4 and described herein are provided as an example only and are not intended to be limiting. Computing device 10, SoC 12, or multi-core processor 14 may include fewer or more than four processor cores 200, 201, 202, 203 illustrated and described herein, individually or in combination. Figure 5 illustrates SoC 12 configured to split the cache memory into component caches in a self-adaptable manner. The SoC 12 can include a variety of components as described above. Some of these components and additional components can be used to implement component caches. For example, the SoC 12 configured to implement component caches can include a system hub 300, a system cache 302, a system cache controller 304, a CPU cluster 306, a protocol converter 308, a GPU 310, a data machine DSP 312, The application DSP 314, the memory interface 316, the camera subsystem 318, the video subsystem 320, the display subsystem 322, the system on-chip network (NoC) 324, the memory controller 326, and the random access memory (RAM) 328. System hub 300 is a component of SoC 12 that manages CPU cluster 306 access to various memories via protocol converter 308 and GPU 310, data processor DSP 312, and application DSP 314. As illustrated in FIG. 5, system hub 300 further manages camera subsystem 318, video subsystem 320, and display subsystem 322 for accessing various memory devices via system No. Each of the subsystems can access the system cache 302 via various paths established by the system NoC 324, the memory interface 316, and the system cache controller 324. A plurality of instances of RAM 328 are coupled to system cache 302 via memory controller 326 and system cache controller 324. Similarly, each of GPU 310, modem DSP 312, and application DSP 314 can access system cache 302 via memory interface 316 and system cache controller 324. In one aspect, system hub 300 can manage system cache 302 and access to RAM 328 to SoC 12. Some of the processors that can access various memories can be included in the CPU cluster 306 and various subsystems (such as camera subsystem 318, video subsystem 320, and display subsystem 322), and can also include other specialized processors. Such as GPU 310, data machine DSP 312, and application DSP 314. System cache 302 can be a shared memory device in SoC 12 that can be used to replace or supplement cache memory that can be associated with various processors and/or subsystems. The system cache 302 can centralize the memory resources of the SoC 12 so that various processors and subsystems can access the system cache 302 to read and write programs designated for repeat and/or fast access. Orders and information. System cache 302 can store various memory devices (such as main memory (not shown), RAM 328, or other storage devices (eg, solid state memory modules, hard disk drives) from various processors and subsystems, as well as from computing devices. ))data of. In one aspect, if a cache miss occurs because the item requested by the system cache 302 cannot be located, the system cache 302 can be supplemented by such memory and storage. In one aspect, system cache 302 can be used as a scratchpad memory for various processors and subsystems. The system cache 302 can be smaller in storage space and entity size than the local cache memory of a SoC having a similar architecture that does not use the system cache 302. However, as further described herein, management of system cache 302 may allow for greater energy savings and equal or better performance speeds of SoC 12, although the system cache has less storage space and physical size. The system cache 302 is logically managed in a partition called component cache. Each component cache contains a limited set of cache lines associated with each other by the component cache identifier. A processing program or thread that is active on computing device 10 can identify component caches when communicating with system cache controller 304. These changes can be made by using a preset component cache identifier for all cached transactions that are received by the system cache controller 304 without a component cache identifier, such as a cache transaction sent by the CPU. Assign to the same component cache. After the component cache is initiated, the cache line belonging to the component cache is allocated according to the demand based on the current workload on the computing device 10. The system cache controller 304 traces the component usage or occupancy by the count of the cache line assigned to the component cache. The cache controller 304 applies or assigns the target occupancy of the component cache to an integer number of cache lines. The respective target occupancy numbers are stored in the scratchpad for use by the system cache controller 304. The component cache is in an overflow state when the component cache exceeds the target occupancy. Component caches can continue to grow by adding additional cache lines to the component cache, but identify the oldest cache line stored in the component cache as obsolete. Therefore, the overflow component cache will have at least two parts: a newest part and an outdated part. Under some conditions, even the latest part may exceed the target occupancy. The identification of obsolete cache lines or obsolete handlers is used by the replacement strategy to identify candidate cache lines for removal from component cache. When a conflict omission occurs, the system cache controller 304 applies a replacement policy in response to the cache allocation request. The system cache controller 304 and the method for adaptively splitting the system cache 302 assign each cache line to a group called a memory bank. Assign a fixed number of banks to the component cache. For the purposes of illustration and explanation, the integer 16 is used. It should be understood that the number of reservoirs may be less than or greater than 16, and this selection may be based on the projected workload of computing device 10 or in accordance with other criteria. Assign a unique storage indicator to 16 storages. In an example embodiment, the bank index is identified by the numbers 0 through 15, which may be represented by 4 bits. This is a lower number of bits than the prior art that can use a coarse time stamp. In this prior art, a much larger number of bits must be used to reduce the occurrence of age omission calculations caused by aliasing. Current circuits and modules use new mechanisms based on outdated bit conditions to achieve the same goal at significantly lower hardware costs. An example method for adaptively partitioning a shared cache further includes the step of assigning a bank indicator and a bank level count to at least one of an integer number of component caches, the integer number of component caches being logically divided into An integer number of banks. For a cache line allocation operation that addresses a particular component cache, increments the bank level value assigned to the bank level count for a particular component cache. When the storage level value reaches a fixed rate of the target occupancy of the respective component cache, the storage indicator is incremented and the storage level is reset. The storage metrics associated with the assigned cache line are stored in the shared cache. In addition, for addressing the individual component cache access operations, the system cache controller 304 identifies a set of candidate cache lines for replacement, and for comparing the candidate cache lines for the set of candidate cache lines The values in the associated storage indicator and the current storage values cached by the individual components are used to calculate the respective ages. The system cache controller 304 determines the outdated bit conditions depending on the respective ages, and stores the obsolete bit conditions in the shared cache. The set of parameters associated with the individual component caches includes the least recently used parameters that conditionally identify when the cache line from the older storage should be moved to the current storage. When the cache line moves from the older storage to the current storage, the system cache controller 304 sets or confirms the new bit. The system cache controller 304 conditionally updates the bank index, the outdated bit condition, and the bank level value of the individual component cache for the component cache access operation that caused the hit condition. When the least recently used parameter is set or confirmed and the storage index is not equal to the component cache current storage index, the new storage index is reset, the definite obsolete bit condition is reset or revoked, and the storage level value is incremented. The system cache controller 304 conditionally identifies replacement candidates for component cache access operations that cause missed conditions. More specifically, system cache controller 304 assigns priority parameters to component caches and conditionally identifies replacement candidates when no obsolete candidates exist. In a preferred sequence, including from any outdated cache line, from the lowest priority cache line having a lower or equal priority than the incoming cache line, and when multiple consecutive replacement candidates are identified, Replace the candidate with the oldest cache line in the shared cache. The system cache controller 304 applies an alternate replacement candidate identification criterion when a plurality of replacement candidates are identified. In an example embodiment, the alternate replacement candidate criteria includes identifying a preset path in the shared cache 302. The system cache controller 304 also assigns optional allocation parameters to individual component caches in the shared cache 302 and conditionally allocates cache lines when the identified component cache is not full. Alternatively, system cache controller 304 does not allocate a cache line when the respective component cache is full and the optional allocation parameters are set or confirmed. 6 is a schematic diagram illustrating the components of the system cache controller 304 of FIG. 5 that enable adaptive segmentation of the system cache 302 in communication with the system cache controller 304. Conventional buffers and logic circuits for enabling communication between various processors on the SoC 12 and the system cache 302 are well known to those skilled in the art. These conventional components are omitted for ease of illustration and description of the adaptive segmentation cache controller and related functions. A system cache controller 304, which in the illustrated embodiment is an adaptive segmentation cache controller, includes an input/output (I/O) port 610 coupled to a communication bus, the bus bar memory interface 316 The memory controller 320 and the protocol converter 308 are coupled to the system cache controller 304. Received on the communication bus 610 from the CPU cluster 306, the camera subsystem 318, the video subsystem 320, the display subsystem 322, the GPU 310, the modem DSP 312, or the application DSP 314 on the execution resources of the thread or handler Cache initialization and allocation requests. As illustrated, the I/O port connector 610a communicates the incoming CCID to both the component cache register 620 and the select logic module 640. The update logic module 650 receives the incoming CCID from the component cache register 620. In an alternative embodiment (not shown), the update logic module 650 can be configured with a dedicated connector to receive the incoming CCID. In addition to the incoming CCID, the cached address is also communicated along the connector 610b to the tag comparison and obsolete logic module 630. As indicated in FIG. 6, there is an instance of the tag comparison and obsolete logic module 630 for each identified path in the shared or system cache 302. As indicated in Figure 6, the I/O port 610 of the adaptive split cache controller 304 receives the incoming transaction address and the associated component cache identifier (or CCID). The CCID is communicated to a set of component cache registers 620 and update logic modules 650. The component cache register 620 includes storage locations for cache configuration parameters, component cache occupancy parameters, and storage parameters. The instance component cache parameters are presented in Table 1. One of the important aspects of the adaptive segmentation cache controller is the ability to define different behaviors of different component caches in the shared cache 302 via settings such as parameters defined in Table 1. New features can be added and the old features can be adjusted for each component cache as needed without affecting the component cache client or SoC infrastructure. The instance component cache occupancy parameters are included in Table 2. Example storage or timing parameters are presented in Table 3. The address identifying a portion of the system cache 302 is forwarded to the tag comparison and obsolete logic module 630 and the selection logic module 640. The tag comparison and obsolete logic module 630 associates tag information with cache lines stored in the system cache 302 or applies tag information to each cache line. The instance tag information is presented in Table 4. Table 1 Table 2 table 3 Table 4 For each CCID, a set of signals is transmitted from component cache register 620 to tag compare and obsolete logic module 630a on bus 625. The signals include current storage metrics, priority levels, LRU bits, and overflow and active condition bits. As indicated, the current storage metrics and priority information are communicated via a plurality of connections of the bus 625. The tag comparison and obsolete logic modules 630a through 630p use various inputs to generate a measure of the age or elapsed time that the cache line has been stored in the system cache 302 to determine if it is out of date. Once the cache is out of line, it remains out of date and the age calculation has no effect. The only case where the reversible cache line obsolescence condition is available is when the cache line is accessed on the read hit and the LRU parameter of the associated component cache is true. In addition, the tag comparison and obsolete logic modules 630a through 630p generate hit, renew, and obsolete condition bits in response to current conditions in the system cache 302. The regeneration and hit status information is forwarded to the update logic module 650. As further indicated in FIG. 6, valid and obsolete conditions along with the component cache priority, component cache identifier, and age indicator are communicated to the selection logic module 640 along the bus 637 for each lane. The details of one of the items of the tag comparison and obsolete logic module 630 and the operation of one of the obsolete programs are illustrated and described in association with FIGS. 7 and 8, respectively. As further indicated in FIG. 6, signals on bus 637 are received by selection logic module 640, which is configured to identify for use with priority level, relative age, CCID, and valid and obsolete condition indicators. The system cache 302 replaces the victim or candidate. As shown in the illustrated embodiment, the selection logic module 640 forwards the identified or victim CCID and the valid victim and victim presence indicator, hit and renew condition indicators, each of which is communicated to the update logic mode. Group 650. The update logic module 650 is configured to apply the configuration parameters along with the self-mark comparison and the hit and re-condition indicators received by the obsolete logic module 630 to update the victim associated component cache and the incoming transaction associated component cache. Occupancy parameters, including current storage level and current storage metrics. FIG. 7 is a diagram illustrating an example embodiment of the tag comparison and obsolete logic module 630 of FIG. The tag comparison and obsolete logic module 630 includes an SRAM 710 that stores tag information, CCIDs, bank indicators or identifiers (BIDs), and valid and obsolete bits of each cache line written to the system cache 302. The multiplexer 720 can controllably transfer the priority level based on the CCID. The comparator 730 generates an input to the AND gate 732 in response to the flag information and the selected address bit received from the input/output port 610. The second input of the AND gate 732 is responsive to a valid state condition or a valid bit stored in the SRAM 710. The output of the AND gate 732 identifies the hit (HIT) condition (i.e., the valid data is already in the cache 302). The multiplexer 722 forwards the current bank indicator or identifier in response to the component cache identifier stored in the SRAM 710. The current bank identifier is coupled to the comparator 740, the BID storage area in the SRAM 710, and the X input of the analog to digital comparator 750. The previously stored BID in SRAM 710 is coupled to the Y input of comparator 740 and analog to digital comparator 750. The output of the analog to digital comparator 750 provides a measure or age at which the cache line has been present in the cache. This metric forwards comparator 752, which identifies when the age is greater than or equal to an integer eight. The output of comparator 752 is coupled to the first input of AND gate 754. The multiplexer 724 forwards the current overflow condition status as directed by the CCID to the second input of the AND gate 754. An OR gate 756 receives from the output of the AND gate 754 the current obsolete bit stored in the SRAM 710 and the inverted version of the current active state as transferred by the multiplexer 726, multiplexed The 726 also operates in accordance with the active indicator of each respective CCID. When the undo start (by undoing the validation indicator) cache component, then all cache lines associated with the cache component become obsolete. The output of OR gate 756 is coupled to the first input of AND gate 758. The AND gate 734 receives a hit (HIT) status signal from the output of the AND gate 732 and further receives information regarding the mode of operation of the currently identified component cache at the second input of the AND gate 734. As illustrated in Figure 7, multiplexer 733 can controllably communicate signals identifying when the corresponding component cache is operating in the least recently used or LRU mode of operation. The output of the AND gate 734 is coupled to the inverting input of the AND gate 758 and the first input of the AND gate 760. The output of the AND gate 758 can controllably update the obsolete or obsolete bits stored in the SRAM 710. Comparator 740 identifies when the current bank index forwarded from multiplexer 722 is not equal to the bank indicator stored in SRAM 710. The output of comparator 740 is coupled to the first input of OR gate 742. The second input of OR gate 742 is coupled to the obsolete bit indicator in SRAM 710. The OR gate 742 will send a logic high to the second input of the AND gate 760 when the bank indicator does not match or when the outdated bit or condition is confirmed, which produces a renew condition signal. As indicated in the flag, when the renew condition signal is true, the bank indicator stored in SRAM 710 is renewed or updated. 8 is a timing diagram illustrating an example embodiment of an obsolete program implemented by the update logic module of FIG. The first assigned cache line is assigned to the first bank (eg, bank 0) according to the obsolete program. When the amount of allocated cache lines is equal to the selected rate of the target occupancy number, the subsequent allocation cache line is assigned to the next bank (eg, bank 1). This procedure continues as the system cache 302 is being filled in by the data. When the amount of allocated cache lines is equal to the target occupancy number, subsequent cache lines are assigned to the next bank, and the cache line in the first bank can be made "outdated". When a new or latest cache line is assigned to the next subsequent bank indicator, the cache line storing the second bank can be made obsolete. The current storage is defined as the storage that is being assigned the incoming allocation. In the steady state, when the reservoir index 0 to 15 is used, the reservoir index K will vary according to 0 <= K <= 15. Assign an incoming cache line to the bank K. Both the storage index K and the CCID store tag information describing the conditions in the shared cache 302 of the storage location in the adaptive segmentation cache controller 304. In the illustrated embodiment, N = 8, M = 2, and 16 is the total number of banks being used by cache controller 304. The variable N is an integer representing the number of the latest banks in the component cache. In such cases, any cache line in a bank having an indicator equal to X may be obsolete when (K - X)%16 >= N (where % represents a modulo operation). When a certain amount of cache line equal to 1/N of the component cache target occupancy has been assigned to the bank K, the current bank index is incremented by setting the bank index to (K + 1)%16. As indicated, the obsolete program identifies candidates for replacement. When a conflict omission occurs, the replacement candidate is selected among the obsolete cache lines, regardless of which component cache has been associated with the outdated cache line. If no outdated cache line candidates exist, the replacement candidates are selected according to the priority scheme depending on the component cache parameters. Immediately after each cache access, all valid cache lines in the set are checked for indications of obsolete or obsolete conditions. The bank ID X of the cache line being checked is compared with the current bank index K of the component cached with the cache line. If the distance (K - X)%16 is greater than or equal to 8, the cache line is identified as obsolete. Under the two conditions, the result of the comparison is ignored. In the first condition, the obsolete cache line will remain out of date, regardless of the result of the comparison. In the second condition, only the cache line that belongs to the overflow component cache is out of date. If the touch or access cache line is assigned to a collection, the cache line is identified as obsolete. Therefore, the obsolescence intensity or frequency will depend on the cache activity. In the timing diagram illustrated in Figure 8, there are 16 banks identified by the storage index 0 to 15 across the abscissa. The reservoir contains a cache line with the same reservoir index. As described, each bank is populated with a target capacity / 8 cache line, where "/" is the division operator. For a newly initiated cache, as indicated in the set 0, which is increasing in time with the number of sets under the ordinate, the allocated cache line is currently filled in the store 0. As also described, once the occupancy of the bank 0 reaches 1/8 of the target occupancy of the bank, the subsequent cache line is assigned to the bank 1 and the cache line in bank 0 is considered as the latest. of. This accumulation or growth continues until the reservoir 7 reaches the current occupancy of 1/8 of the target occupancy of the storage 7. At that time, the cache line in banks 1 to 7 is regarded as the latest, the subsequent allocation is assigned to the bank 8, and the cache line in the bank 0 can be made obsolete. When the bank 8 reaches 1/8 of its component cache target occupancy, the cache line in the banks 2 to 8 is regarded as the latest, the subsequent allocation is assigned to the bank 9, and the bank 0 and the bank can be stored. The cache line in body 1 is out of date. When the cache has been active for a long time and sufficient activity is received to fill the bank 15 to 1/8 of its target occupancy, the unfrozen cache line in bank 0 will be up to date. The newly allocated cache line can be assigned to bank 0 and the cache line in bank 1 to bank 8 can be made obsolete. When this is the case, the cache line 0 and the cache line in the reservoir 9 to the reservoir 15 are considered to be up-to-date. When the bank 0 reaches 1/8 of the target occupancy, the subsequent cache line is assigned to the bank 1, and the cache line in the bank 2 to the bank 9 can be made obsolete. When this is the case, the cache lines in the reservoir 0, the reservoir 1 and the reservoir 10 to the reservoir 15 are considered to be up-to-date. The above method can be generalized into an array of N x M reservoirs in which there are N most recent reservoirs and (N - 1) x M obsolete reservoirs, where N and M are integers. In the steady state, the relationship is such that for the current bank K, 0 <= K <= (N*M - 1), when (K - X)%(N*M) >= N, the bank ID can be made The cache in X is out of date. When a certain amount of cache line equal to 1/N of the component cache target occupancy has been assigned to the bank K, the current bank index is set to (K + 1)% (N*M). The simulation shows that N = 8 and M = 2 provide excellent performance, and larger values will more or less improve performance at the expense of more memory for storing information on the memory of each cache line. FIG. 9 is a diagram illustrating an example embodiment of the selection logic module 640 of FIG. 6. The selection logic module 640 receives the CCID, the priority identifier bit, and the time or age measure of the selected cache line in the system cache 302 and the obsolete, valid, renewed, and hit status indications for each candidate cache line. symbol. In response, the selection logic module 640 responds to the presence or absence of the identified victim and conditionally generates or identifies the victim cache line and the control signal. The CCID communicated from the tag comparison and obsolete logic module 630 is applied to the multiplexer 950 and the comparators 902a through 902p. Comparators 902a through 902p further receive the incoming CCID associated with the incoming transaction and identify which of the information vectors labeled "Same CCID Vector" is forwarded to the first set of inputs of multiplexer 918 when the match exists. Comparators 904a through 904p are configured to receive priority information forwarded from the tag comparison and obsolete logic module 630 and provide respective output signals when a medium level of the three priority levels is present. The resulting information vector labeled "Medium Priority CCID Vector" is forwarded to the second set of inputs of multiplexer 918. The OR logic gate 914 forwards the logic high signal to the first input of the AND gate 916 when the CCID having the medium level priority exists. When the incoming CCID has a high priority (as indicated by the confirmed signal on the connection marked "High Priority Incoming CCID" at the second input of the AND gate 916), the "AND" gate 916 is confirmed. The output, which directs the multiplexer 918, forwards the medium level priority CCID vector to its output. Otherwise, when the output of the AND gate 916 is not logic high, the multiplexer 918 forwards the same CCID information vector to its output. The output of multiplexer 918 is forwarded to the first set of inputs of multiplexer 924. Comparators 906a through 906p are configured to receive priority information forwarded from the tag comparison and obsolete logic module 630, and provide separate levels when lower levels of the three priority levels exist at the output of comparators 906a through 906p output signal. The resulting information vector labeled "Low Priority CCID Vector" is forwarded to the second set of inputs of multiplexer 924. The OR logic gate 920 forwards the logic high signal to the first input of the AND gate 922 when the CCID having the lower priority is present. When the incoming CCID has a medium or low priority (as indicated by the confirmed signal on the connection marked "High or Medium Incoming CCID Priority" at the second input of the AND gate 922), "and" The output of the gate 922 directs the multiplexer 924 to forward the lower priority CCID vector to its output. Otherwise, when the output of the AND gate 922 is not asserted, the multiplexer 924 forwards the output of the multiplexer 924 to its output, the output of which is coupled to the input of the selected oldest candidate logic module 926. The selected oldest candidate logic module 926 further receives the age value from the individual tag comparison and obsolete logic module 630 on the age indicator input, and uses the respective age information to determine the location as received from the multiplexer 924. Which of the cache candidates is the oldest. The resulting information vector is forwarded to the first set of inputs of multiplexer 930. Passing all obsolete candidates identified by the vector labeled "outdated vector" to the input of OR gate 928 and the second set of inputs of multiplexer 930, when at least one of the passage-based obsolete indicators is confirmed, The vector is forwarded to the output of multiplexer 930 as controlled by the output of OR gate 928. Otherwise, when no candidate is identified as obsolete, multiplexer 930 forwards the resulting information vector identifying the oldest cached information to the first set of inputs of multiplexer 934. Each of the path-based valid indicators received from the tag comparison and the obsolete logic module is applied at the input of the respective inverters 932a through 932p. The respective outputs of inverter 932 are applied at the input of OR gate 933 and provide an "invalid vector" for the second set of inputs of multiplexer 934. The OR gate 933 identifies when any of the candidates is invalid, and when this is the case, the candidate received at the second set of inputs of the multiplexer 934 is forwarded to its output. Otherwise, when no candidate is identified as invalid, the output of OR gate 938 directs the output of multiplexer 930 to its output. The output of multiplexer 934 is forwarded to the first set of inputs of multiplexer 942. Each of the path-based renewing indicators received from the tag comparison and the obsolete logic module is applied at the input of the OR gate 940. The output of the OR gate 940 is coupled to the control input of the multiplexer 942 and provides a renewed output signal that is forwarded to the update logic module 650. When the renew signal is asserted, the renewed information vector received at the first set of inputs of multiplexer 942, labeled "Renew Vector", is forwarded to the input of any logic module 946. Otherwise, multiplexer 942 forwards the output of multiplexer 934 to any of the logic modules 946 when there is no new signal confirmation. In a preferred embodiment, any logic module 946 is selected to select the leftmost path with the selected candidate, which is applied to the control input of multiplexer 950. When the renew condition is confirmed, the victim CCID at the output of the multiplexer 950 corresponds to the CCID of the addressed cache line. The new condition indicates that the update logic 650 updates the bank parameters. As further indicated in FIG. 9, each of the path-based hit condition indicators received from the tag comparison and obsolete logic module 630 is applied at the input of the OR gate 944. The output of the OR gate 944 provides a hit condition output signal that is forwarded to the update logic module 650. FIG. 10 is a flow diagram illustrating an example embodiment of a method 800 for adaptively partitioning a shared or system cache 302. The method 800 begins at block 802 where parameters (including identifiers, target occupancy, occupancy counts, and memory metrics) defining a selected number of components of a shared cache are stored (such as) a set of component caches. The memory 620 and/or the synchronous random access memory. As shown in block 804, when various client terminals of the shared cache 302 receive a system cache request, the adaptive segmentation cache controller 304 uses the stored parameters and the current cache condition for cache allocation. When a cache line is allocated, the bank metric is associated with the cache line and the count is incremented. At the same time, or as early as possible, as indicated in block 806, the adaptive segmentation cache controller 304 uses the stored current bank index and occupancy count to identify the cache line for possible self-cache retrieval. As further indicated in block 806, the associated occupancy count is decremented when the cache line is retracted or the cache line is removed from the cache. As shown in block 808, under certain conditions, the adaptive segmentation cache controller 304 permits the identified component caches to exceed the respective assigned target occupancy. FIG. 11 is a schematic diagram illustrating an example embodiment of computing device 1100. In the illustrated embodiment, computing device 1100 includes a processor 14 communicatively coupled to processor readable medium 1200, system or shared cache 302, and adaptive segmentation cache controller 304. The adaptive split cache controller 304 communicates with the processor 14 and the system or shared cache 302 via the input/output port 610 and the bus bar 1120. As further shown in FIG. 11, the adaptive segmentation cache controller 304 includes a component cache register 620 for storing configuration parameters, timing parameters, and occupancy parameters. Information that is forwarded and re-stored in component cache register 620 based on the signals communicated on bus 1120. The tag comparison and obsolete logic module 630 receives the information stored in the component cache register 620 and the address information from the client request communicated from the processor 14. The tag comparison and obsolescence logic module 630 forwards a set of signals to the selection logic module 640 for each path that identifies one or more obsolete partitions for removal from the shared cache 302. The tag comparison and obsolete logic module 630 also generates a renew and hit signal and communicates it to the update logic module 650. The update logic module 650 updates the information in the component cache register 620 using the signals and status information about the potential victim for removal, and removes the cache line from the shared cache 302 according to the selection criteria. / or adjust the partition in the shared cache 302. The computing device 1100 is capable of splitting the system shared cache 302 when configured with the adaptive split cache controller 304 as illustrated and described such that the individual partitions are independent of the set and path structures. As shown in the illustrated embodiment, system shared cache 302 includes three identified component caches. The first component cache identifier CCID 0 has a target occupancy of 16 cache lines, but has utilized the current cache condition and the workload on the computing device 1100 to currently occupy 22 cache lines. The second component cache identifier CCID 1 has a target occupancy of two, but has utilized the current cache condition and the workload on the computing device 1100 to occupy 4 of the set of available cache lines. In addition, the third component cache identifier CCID 2 has a target occupancy of eight cache lines and currently uses eight cache lines. FIG. 12 is a diagram illustrating an example embodiment of a processor readable medium 1200 stored in memory 16 in a computing device. As indicated, the processor readable medium 1200 includes a set of storage locations and logic modules or executable instructions that, when executed by the processor 14, perform and are used to adaptively partition and manage the shared or system cache 302. The various steps associated with outdated, updated, and selected procedures are described. As indicated in the illustrated embodiment, the storage location includes a component cache configuration parameter 1210, an occupancy number parameter 1220, and a time related parameter 1230. As also indicated, the logic module includes shared cache access identification logic 1240, memory identification logic 1250, and replacement candidate selection logic 1260. The shared cache access identification logic 1240 identifies the client cache request. The bank processing logic 1250 assigns a bank metric to the cache line and tracks the current component cache occupancy and the metric of the time that the cache line already exists in the cache. The replacement candidate selection logic 1260 identifies one or more candidates for removal from the cache in a prioritized scheme in response to the use of the operational parameters and cache. As described, the storage identification logic 1250 and the replacement candidate selection logic 1260 will store the updated information in the occupancy parameter storage area 1220 and the time dependent parameter storage area 1230. In order for a computing device to function as described, certain steps in the procedures or processes described in this specification naturally precede other steps. However, if the order of the steps described does not alter the described functionality, the computing device and the methods described are not limited to the order or sequence. That is, it has been recognized that some steps can be performed before, after, or in parallel with (substantially synchronized with) other steps without departing from the scope of the claimed computing device and method. In some cases, certain steps may be omitted or not performed. In addition, words such as "subsequent", "continued", "next", "subsequent", etc. are not intended to limit the order of the steps. These terms are only used to guide the reader through the description of the exemplary methods. In addition, a person of ordinary skill in the art of energy management in a portable, desktop or server computing device can identify appropriate hardware and/or circuitry and/or identify appropriate logic and decisions for implementation without difficulty based on the present specification. The disclosed embodiment. Thus, the disclosure of a particular set of code instructions, decision thresholds, or detailed hardware devices is not deemed necessary for a thorough understanding of how to make and use the disclosed computing devices and methods of operation. The inventive functionality and aspects of the program and circuit architecture claimed by the processor are explained in greater detail in the above description and in conjunction with the drawings. FIG. In one or more of the illustrative aspects as indicated above, the functions described may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functions may be stored as one or more instructions or code on a computer readable medium, such as a non-transitory processor readable medium. Computer readable media includes data storage media. The storage medium can be any available media that can be accessed by a computer or processor. By way of example and not limitation, such computer-readable media can comprise RAM, magnetic RAM (MRAM), ROM, EEPROM, NAND flash memory, NOR flash memory, spin torque MRAM, CD-ROM or Other optical disc storage, disk storage or other magnetic storage device, or any other medium that can be used to carry or store the desired code in the form of an instruction or data structure and accessible by a computer. As used herein, magnetic disks and optical disks include compact discs ("CDs"), laser compact discs, optical compact discs, digital versatile compact discs ("DVD"), flexible magnetic discs and Blu-ray discs, where the magnetic discs are usually magnetically The data is reproduced, and the optical disk optically reproduces the data. Combinations of the above should also be included in the context of non-transitory computer readable media. While the invention has been described and illustrated in detail, it is understood that various modifications and changes can be made herein without departing from the scope of the invention.

1‧‧‧Cache storage area
2‧‧‧Cache partition
3‧‧‧Cache partition
4‧‧‧Cache partition
5‧‧‧Cache storage area
6‧‧‧Cache partition
7‧‧‧Cache partition
8‧‧‧Cache partition
10‧‧‧ Computing device
12‧‧‧System Single Chip (SoC)
14‧‧‧ Processor
16‧‧‧ memory
18‧‧‧Communication interface
20‧‧‧ Storage interface
22‧‧‧Communication components
24‧‧‧Storage components
26‧‧‧Antenna
28‧‧‧Network interface
30‧‧‧Wireless network
32‧‧‧Wireless connection
40‧‧‧Internet
44‧‧‧Wired connection
50‧‧‧ remote computer
200‧‧‧ processor core
201‧‧‧ processor core
202‧‧‧ Processor Core
203‧‧‧ processor core
300‧‧‧System Hub
302‧‧‧System Cache/Share Cache
304‧‧‧System cache controller
306‧‧‧Central Processing Unit (CPU) Cluster
308‧‧‧Commitment Converter
310‧‧‧Graphical Processing Unit (GPU)
312‧‧‧Data machine digital signal processor (DSP)
314‧‧‧Application Digital Signal Processor (DSP)
316‧‧‧ memory interface
318‧‧‧ camera subsystem
320‧‧‧Video subsystem
322‧‧‧Display subsystem
324‧‧‧System on-chip network (NoC)
326‧‧‧ memory controller
328‧‧‧ Random Access Memory (RAM)
610‧‧‧Input/Output (I/O)埠
610a‧‧‧Input/Output (I/O)埠 Connector
610b‧‧‧Connecting parts
620‧‧‧Component cache register
625‧‧ ‧ busbar
630‧‧‧Marker comparison and obsolete logic module
630a‧‧‧Marker comparison and obsolete logic module
630b‧‧‧Marker comparison and obsolete logic module
630p‧‧‧Marker comparison and obsolete logic module
637‧‧‧ busbar
640‧‧‧Select logic module
650‧‧‧Update Logic Module
710‧‧‧SRAM
720‧‧‧Multiplexer
722‧‧‧Multiplexer
724‧‧‧Multiplexer
726‧‧‧Multiplexer
730‧‧‧ Comparator
732‧‧‧"and" gate
733‧‧‧Multiplexer
734‧‧‧"and" gate
740‧‧‧ comparator
742‧‧‧"or" gate
750‧‧• Analog Comparator
752‧‧‧ Comparator
754‧‧‧"and" gate
756‧‧‧"or" gate
758‧‧‧"and" gate
760‧‧‧"and" gate
800‧‧‧ method
802‧‧‧ block
804‧‧‧ Block
806‧‧‧ Block
808‧‧‧ Block
902a‧‧‧ comparator
902p‧‧‧ comparator
904a‧‧‧ comparator
904p‧‧‧ comparator
906a‧‧‧ comparator
906p‧‧‧ comparator
914‧‧‧"or" logic gate
916‧‧‧"and" gate
918‧‧‧Multiplexer
920‧‧‧"or" logic gate
922‧‧‧"and" gate
924‧‧‧Multiplexer
926‧‧‧Selected oldest candidate logic module
928‧‧‧"or" gate
930‧‧‧Multiplexer
932a‧‧‧Inverter
932p‧‧‧Inverter
933‧‧‧"or" gate
934‧‧‧Multiplexer
938‧‧‧"or" gate
940‧‧‧"or" gate
942‧‧‧Multiplexer
944‧‧‧"or" gate
946‧‧‧Select any logic module
950‧‧‧Multiplexer
1100‧‧‧ Computing device
1120‧‧‧ busbar
1200‧‧‧Processable media
1210‧‧‧Component cache configuration parameters
1220‧‧‧ occupancy parameter
1230‧‧‧ time related parameters
1240‧‧‧Shared cache access recognition logic
1250‧‧ ‧ storage identification logic
1260‧‧‧Replacement candidate selection logic

BRIEF DESCRIPTION OF THE DRAWINGS The accompanying drawings, which are incorporated in FIG In the drawings, the same reference numerals refer to the For reference numerals having an alphabetic character name such as "102A" or "102B", the alphabetic character name can distinguish between two identical parts or elements present in the same figure. Whenever a reference numeral is used to cover all parts of the drawings having the same reference numerals, the letter character names of the reference numerals may be omitted. Figure 1 is a schematic diagram illustrating the conventional path segmentation of a shared cache. 2 is a schematic diagram illustrating a conventional set of shared caches and path segmentation. 3 is a functional block illustrating an example embodiment of a computing environment including a computing device suitable for implementing an improved cache controller. 4 is a schematic diagram illustrating an example embodiment of a multi-core processor that can be implemented in the computing device of FIG. FIG. 5 is a schematic diagram showing an example embodiment of the system single wafer of FIG. 3. FIG. 6 is a schematic diagram showing the components of the shared cache controller of FIG. 5. 7 is a schematic diagram showing an example embodiment of the tag comparison and obsolete logic module of FIG. 6. 8 is a timing diagram illustrating an example embodiment of an obsolete program implemented by the obsolete logic module and the update logic module of FIG. 9 is a schematic diagram showing an example embodiment of the selection logic module of FIG. 6. 10 is a flow chart illustrating an example embodiment of a method for adaptively partitioning a shared cache. 11 is a schematic diagram illustrating an example embodiment of a computing device. 12 is a schematic diagram illustrating a computing device including an example embodiment of a processor readable medium.

Claims (30)

A method for adaptively splitting a shared cache, the method comprising the steps of: storing parameters defining an integer number of component caches in the shared cache, wherein at least one of the integer number of component caches One is associated with a set of parameters, the set of parameters includes a target occupancy number and a occupancy count; and a unique identifier corresponding to one component cache of the integer number of component caches is applied to the client request; Responding to the client requests for storing the client data in the shared cache, using the information stored in the set of parameters to perform cache line allocation, wherein when a cache line is allocated, an associated occupation is incremented Counting; and using the information stored in the set of parameters to identify candidate cache lines for reclaiming, wherein when a single cache line is reclaimed, the associated occupancy count is decremented. The method of claim 1, further comprising the step of assigning an active indicator to the at least one of the integer number of component caches, wherein the cache lines present in the shared cache are in each of The component cache is associated with an outdated condition by default when the active indicator is revoked. The method of claim 2, further comprising the step of assigning a bank indicator and a bank level count to the at least one of the integer number of component caches, wherein the integer number of components are cached At least one of the at least one is divided into an integer number of banks; and For addressing each cache line assignment of the at least one of the integer number of component caches, incrementing a bank level value in the bank level count for the at least one of the integer number of component caches And when the storage level value reaches a fixed fraction of the target occupancy of the at least one of the integer number of component caches, incrementing the storage indicator and resetting the storage level count; The storage indicator is associated with an allocated cache line for storage in the shared cache. The method of claim 3, further comprising the steps of: identifying an access operation of the at least one of the integer number of component caches, identifying a set of candidate cache lines for replacement; and for the set of candidates Each candidate cache line in the cache line, by comparing a value of the memory indicator associated with a particular candidate cache line with a current memory bank associated with the component of the particular candidate cache line The indicator value is used to calculate an age. The method of claim 4, further comprising the steps of: determining, for each of the candidate cache lines, the obsolete condition depending on the respective age, and associating the obsolete condition with the particular candidate Take a line to store in the shared cache. The method of claim 4, wherein the set of parameters includes a least recently used parameter that conditionally identifies when a cache line from one of the older banks should be moved to the current bank. The method of claim 5, further comprising the steps of: conditionally updating the memory indicator of the accessed cache line and the obsolete condition for a component cache access that causes one of the hit conditions, and conditional Update the storage tier value of the component cache. The method of claim 7, wherein when the component cache recent least used parameter is set and the storage index is not equal to the component cache current storage index, the storage indicator is renewed, the obsolete condition is reset, and the value is incremented. The reservoir level value. The method of claim 2, further comprising the step of conditionally identifying a replacement candidate for a component cache access that causes one of the missing conditions. The method of claim 9, further comprising the steps of: assigning a priority parameter to the at least one of the integer number of component caches; and conditionally identifying the replacement candidate when no obsolete candidate exists By. The method of claim 10, wherein in a preferred sequence, including from any outdated cache line, from the lowest priority cache line having a lower priority than an access cache component, from the location The access cache component is accessed and the replacement candidate is selected from an oldest cache line when a plurality of recent replacement candidates are identified. The method of claim 11, further comprising the step of applying an additional replacement candidate criterion when a plurality of replacement candidates are identified. The method of claim 12, wherein the additional replacement candidate criteria comprises identifying a predetermined path. The method of claim 1, further comprising the steps of: assigning an optional allocation parameter to the at least one of the integer number of component caches; and conditionally allocating a cache line. The method of claim 14, wherein a cache line is not allocated when the respective component cache is full and the selectable allocation parameter is set. An adaptive split cache controller includes: a scratchpad configured to store a set of operational parameters for a desired number of component caches for a shared cache; mark comparison and obsolete logic, a unique component cache identifier coupled to the buffers and configured to receive a cache address and identify one of the shared caches, the flag comparison and obsolete logic responding to an operation The mode input generates a measure of the elapsed time that the information has been stored in a particular cache line of the shared cache, conditionally records an obsolete condition, a valid condition, and updates a storage indicator; and selection logic Connected to the tag comparison and obsolete logic and the registers, the selection The logic is configured to receive the measure of elapsed time for the information stored in the particular cache line in the shared cache, the unique component cache identifier, and a priority indicator, and based on the obsolete condition and The valid condition identifies a component cache identifier of one of the candidate cache lines to be replaced in the shared cache. The adaptive segmentation cache controller of claim 16 wherein the set of operational parameters includes a target occupancy number and a current occupancy number. The adaptive segmentation cache controller of claim 17, wherein the set of operational parameters includes an active indicator, and the cached lines present in the shared cache are cached at their respective components to cause the active indicator When the confirmation is revoked, it is associated with the obsolete condition by default. The adaptive segmentation cache controller of claim 18, wherein the set of operational parameters includes one or more of a priority value, a current storage indicator, a current storage level, and a least recently used parameter. The adaptive segmentation cache controller of claim 16, wherein the selection logic is configured to enable a preferred sequence, including from any outdated cache line, having a lower priority than an access cache component The lowest priority cache line, identifies a replacement candidate from the accessed cache component, and identifies the replacement candidate from an oldest cache line when a plurality of latest replacement candidates are identified. Adapting the cache controller as described in claim 16 wherein the selection logic is configured to root The replacement candidate is identified based on at least three priority levels. A computing device, comprising: a processor; a shared cache configured to provide a data storage resource for the processor; an adaptive split cache controller coupled to the processor in a communication manner The adaptive cache controller further includes: an input port configured to receive a cache access request from a client; a register configured to store a set of operational parameters, The set of operational parameters are used for the desired number of component caches of the shared cache; tag comparison and obsolete logic coupled to the registers and configured to generate one of the pair of information stored in the shared cache Measure one of the elapsed times in a particular cache line, record an outdated condition, and conditionally update a bank indicator; and select logic coupled to the tag comparison and obsolete logic and the register, the selection The logic is configured to conditionally identify one of the replacement candidates to be replaced in the shared cache. The computing device of claim 22, wherein the registers are configured to store a target occupancy and a current occupancy. The computing device of claim 22, wherein the registers are configured to store configuration parameters, occupancy parameters, and time related parameters. The computing device of claim 24, wherein the time dependent parameters comprise a current storage finger Marked with a storage level. The computing device of claim 24, wherein the occupancy parameters include a current occupancy number and an overflow bit. The computing device of claim 24, wherein the configuration parameters are selected from the group consisting of: a target occupancy number, a priority level, an active bit, a least recently used parameter, and an optional allocation parameter. . A non-transitory processor readable medium having stored thereon processor-executable instructions that, when executed, direct a processor and an adaptive segmentation cache controller to perform operations comprising: defining cache parameters, The cache parameters include a unique identifier of a component cache of a shared cache, a target occupancy of the component cache, an occupancy count associated with the component cache, and an integer number of storage indexes. Each storage indicator is associated with one of a corresponding number of component caches, and at least one of the component caches is divided into an integer number of storages; the identification is from a client application to the shared fast Taking a cache access request of a component cache; generating a measure of elapsed time in a particular cache line of the component cached for the component cached in the shared cache; in response to the elapsed time Metrically recording an obsolete condition; conditionally updating a storage indicator for the cache line, the storage indicator identifying a storage in the storage of the component; and Conditionally identifying one of the replacement candidates to be replaced in the shared cache. The non-transitory processor readable medium of claim 28, wherein the metric that generates an elapsed time for the information that has been stored in the particular cache line in the shared cache is in response to a storage level, the storage The body level represents an occupancy fraction of the bank. The non-transitory processor readable medium of claim 28, further comprising: defining a set of operational parameters including an active indicator, wherein the cached lines present in the shared cache are cached in their respective components Associated with the obsolete condition by default when the active indicator is confirmed.